Wealth and possessions; at least until you’ve accumulated so much you couldn’t possibly spend it all in a single day even if you wanted to.
Beyond explicitly countable “stuff”, I’d argue that not only are flavor-text stats more engaging and immersive, they can also be substantially more clear and less ambiguous than pure numerical stats.
Think about it: Is “Strength: 40%” outrageously high for an MC in a given situation, or pathetically low? Is it enough to move a box that’s clearly heavy enough only stronger-than-average humans can move it? What about unbalancing that boulder blocking the mountain pathway? Knowing that you have a certain number doesn’t actually mean much unless you know what you’re being tested against. I can’t actually think of any CS games that say things like
#Try to roll the boulder. [Req. 45 STR]. Sure, by the end of the game you can assume 50% Strength is about average, but it’s nowhere near average at the beginning or even middle of most games. Most games, that’s about the maximum you can get in the middle, and far more than anyone can get at the beginning.
But if you’ve got a flavor-text display for your stat, and you set the steps of that flavor-text display to match up with difficulties of the stat checks you’ll be asking players to perform, they’ll be able to estimate their abilities with great accuracy. Imagine a part of the game where “easy” stat checks are around 20, “moderate” around 30, and “hard” around 40. You can use your flavor text to tell players with 20 “Your [strength/speed/whatever] isn’t very impressive,” while giving players with 40 in the stat something like “Your [strength/speed/whatever] is far superior to most in your circumstances.”
Then in later chapters, as stat checks become more demanding, you can use different thresholds to determine whether players see their stats as “…not particularly impressive at this point,” “…above average for those who have been through what you have,” or whatever, you get the idea so that players can reasonably and reliably gauge their progress against the current difficulty. It may even help avoid situations where a player uses all their stat increases on a single stat out of some fear they’ll fail a check if they spread their stats out evenly.
That leads me to suggest numeric stats have two more useful situations: First, when there’s a large difference between the start and end of a game regarding what’s considered “high” and “low” for a stat. Creating new flavor-texts for your stats every time you change the threshold of “good enough” will get tedious pretty quickly. Second, when an author feels like writing the flavor-text is exactly the wrong kind of work for them to enjoy.